AMD CEO Dirk Meyer: Smart Men Have Foundries

It's been about two years since Advanced Micro Devices started talking about what it called an "asset smart" strategy for organizing its manufacturing assets in a more financially sustainable way. As one of two vertically integrated x86 microprocessor makers along with larger rival Intel, AMD had many in the industry wondering just how far it would go in lightening its cost-intensive manufacturing load.

Last summer, new CEO Dirk Meyer and his team pulled the trigger, sealing a deal to spin off AMD's silicon fabrication operations to Advanced Technology Investment Co., an investment company owned by the government of Abu Dhabi in the United Arab Emirates. The move to form the new entity, called The Foundry Company, was approved by AMD shareholders this month. AMD will own about 35 percent of The Foundry Company. That stake is down from the 45 percent that was originally intended when the deal was struck, due to AMD's continued financial problems and falling share price.

Meyer spoke with Channelweb.com Friday about the spin-off and its implications for AMD's system builder channel. He also touched upon the way forward in a tough market for AMD as a fabless design and marketing house and for The Foundry Company as a semiconductor manufacturer with one very loyal customer, hoping to have more.

So congratulations on the result from your shareholders. Clearly this is something AMD has wanted to do and has planned to do for a long time. But at the risk of going for the joke to start off, what does this all mean in terms of AMD founder Jerry Sanders' famous quote, "Real men have fabs?"

Jerry was speaking at a particular point in time, in the context of being at the start of the semiconductor business in Silicon Valley. But times have changed. The capital expenditures for manufacturing have grown. You are seeing semiconductor companies that used to have fabs start to have foundries and that's only going to accelerate. The formation of The Foundry Company does two things -- one is that we're creating a very technologically capable company and also that it's a financially sound company.

We are already seeing some of our fabless semi partners react very positively to The Foundry Company. Going back to Jerry Sanders, I like to say now that, "Smart men have foundries."

If you own an equity interest in The Foundry Company of about 35 percent, what are your obligations as a public company to report Foundry Company financials or your stake of those financials? And how fast does this all happen?

Our CFO Bob Rivet will be hosting a Webcast next week to explain this more fully to financial analysts. But first of all, AMD will become a product, design and marketing focused company. The company that I manage will become that. It's true that AMD will own a stake in The Foundry Company at the share you stated. And for a period of time we will be obligated to consolidate Foundry Company financials on the AMD balance sheet and P&L statement. For a while, we will be their only customer. But as The Foundry Company grows and gets other customers, we will no longer have to do that. We will have two separate management teams. In terms of cash flows, ours will be used to support product development needs and to procure materials from The Foundry Company, and theirs will be put toward manufacturing and process technology.

What does the spin-off mean in terms of process technology and how quickly The Foundry Company follows the industry leader, whether Intel or some other company?

To give you some history perspective, we have lagged by six to 12 months behind whoever's in front on process technology. In recent years, that's been Intel. The reason for that is twofold, in that it has to do with both scaling concerns and business strategy. It's extremely expensive to ramp the next node and the tools to do so, and it doesn't make sense for us to be line-on-line with Intel in that regard. So we generally start to ramp a new technology when it's more mature. From a manufacturing cost perspective, that's been a good thing for us. And architecturally, we've still been able to be competitive in the market.

But that could change somewhat as The Foundry Company scales, as financial considerations start to change, as they become more economically able to move more aggressively and start to get copies of the latest tools from the leaders sooner, and they could have an opportunity to close that gap.

What does this all mean for your channel?

The headline is that this transaction allows AMD to do two things -- to transform our business model to both a more sustainable and a more focused company. That's beneficial to our channel, too. One thing we're doing is increasing our focus on the channel, and we're committed to our mission to provide our channel partners with the ability to procure and market our platforms. The thing is, ATI had a channel organization and AMD has a channel organization, but it's been tough [since the ATI acquisition] to bring together the channel message we have about integrating the GPU and the CPU most effectively. We're doubling down on that message and that commitment.

Next: Gearing Up For The Future

The last time AMD grabbed a lot of market share from Intel, you seemed to have trouble scaling as a company in terms of product development consistency, you know, to take advantage of that. It's a tough economy, obviously, so it's a little odd to talk about growing right now, but how do you feel AMD is positioned in terms of scalability, with the spin-off and maybe from the lessons learned from Barcelona?

Are you talking about scaling market share or R&D? I'm not sure I follow.

Well, with Opteron, for example, you took market share, but in conversations I've had with [Computing Solutions Group Senior Vice President] Randy Allen and [Server and Workstation Division General Manager] Pat Patla, they indicated that maybe the company wasn't ready with its internal processes to keep producing consistently to take full advantage of the share you gained or to hold onto it. So you have a Barcelona, which was delayed due to a glitch, but then you follow up with Shanghai, which was a success and you actually shipped it ahead of your road map. And Randy and Pat both talked a lot about how AMD made changes to its validation processes to make that happen. So my question is basically how do you make sure you get fewer Barcelonas and more Shanghais going forward, and where do you think AMD's internal development processes are, relevant to what the market may do going forward?

The best answer is really the history since I've been at the company. We started really producing innovation on the x86 platform back in the 1990s, which was just a very interesting and exciting time for us. And we really differentiated ourselves with the Opteron and AMD 64 products, then increased our R&D staff and started developing new products and serving new customers.

And that meant a lot of work and difficult things to do, like getting into the enterprise arena. It really was a challenge. And that helps me to understand why we had a few hiccups on Barcelona. But as I have said before, these are the most complicated devices on the face of the planet, so hiccups happen. Nehalem was on Intel's drawing board for a long time before they got it out, though I don't want to make this a finger-pointing session about them. We have our R&D staff in place, and the economy being what it is, we've taken a pause in growing that R&D staff. But over the last year-and-a-half, we've been working on our product development team to make it ready for a potentially larger company.

Let's talk about Fusion, which is actually now your tagline, "The future is Fusion." You've been stressing the integration of the GPU and CPU on a platform, like with the Dragon platform you introduced in January. But where do your Fusion plans -- the full integration of the CPU and GPU on a single die or in a single package -- fit into this new arrangement for AMD?

There are two ways to answer the question, the first being physical chip partitioning in one package or one piece of silicon, which makes sense for some product economics. But the more interesting element to me is this idea of what processing engines are optimal for application workloads to actually run on. The traditional wisdom is that the GPU just does the graphics rendering problem and the CPU does everything else. But what we've seen actually happening is that graphics processors are taking on more general workloads. We look at that in the future as something we call "compute offload," or "accelerated computing." Or in the nearer term, we've called it "stream computing." The fact is, the GPU has far more computational capability than the CPU in certain key areas and it's more energy efficient.

I saw you quoted as saying that the term "netbook" will become meaningless in a year or so. Can you elaborate on that?

I think that the netbook today doesn't represent a new category. It's just a smaller form-factor notebook, so we're just seeing a continuum of notebook form-factors. And we can see full-size notebooks at the high end of the price point for netbooks. I wonder why anyone would want a netbook when they could have a much more powerful notebook for roughly the same price. And when the buzz cools off a bit, I think we'll just be calling all these things "notebooks" again. That said, it's good to see some innovation in form-factors. One thing that's always amazed me personally is that for years, there's been so little innovation in the PC industry, so little innovation in terms of industrial design, and it's been boring. But I think we will see increasing innovation in industrial design, especially in desktop designs, with smaller footprint PCs and specialty desktops and so on.